🚀 go-pugleaf

in article <76000295@p.cs.uiuc.edu>, gillies@p.cs.uiuc.edu says:
> Nf-ID: #R:bruce.oz:578:p.cs.uiuc.edu:76000295:000:833

> - The 68851 is a SLOW paging chip -- it adds 1 wait-state to A/UX memory
>   references.  Sun/etc. get around this with their own PMMU designs.  Does
>   the 68030 have this wait state?  If not, then putting the 68030 in
>   the IIx speeds up A/UX (but not the Mac O/S).

The 68030 MMU does its translations fully in parallel, so as long as your 
translation is in the ATC, or the MMU's shut off, no delay results.  The '851
adds a wait state whether it's translating or not.  On the other hand, the
'851 ATC is 64 entries, whereas the '030 ATC is only 22 entries.

Whether or not the MacIIx speeds up memory-cycle-wise over the MacII really
depends on where the MacII's extra wait state came from.  If it really was
from the MMU alone, it should go one cycle faster.  If it was an artifact of
the memory system design, no change.  The 15% speeup they quote sounds to me
more like a data cache speedup; does the normal Mac OS permit the data cache
to be enabled?
-- 
Dave Haynie  "The 32 Bit Guy"     Commodore-Amiga  "The Crew That Never Rests"
   {ihnp4|uunet|rutgers}!cbmvax!daveh      PLINK: D-DAVE H     BIX: hazy
		"I can't relax, 'cause I'm a Boinger!"